|Publication number||US6411603 B1|
|Application number||US 09/122,032|
|Publication date||Jun 25, 2002|
|Filing date||Jul 23, 1998|
|Priority date||Jul 23, 1998|
|Publication number||09122032, 122032, US 6411603 B1, US 6411603B1, US-B1-6411603, US6411603 B1, US6411603B1|
|Inventors||Sudhir R. Ahuja, Ioannis A. Korilis|
|Original Assignee||Lucent Technologies Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (7), Referenced by (27), Classifications (19), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
The invention relates to techniques for management of a network, and more particularly to a technique for allocating limited capacity of such a network for utilization by multiple users in accordance with a pricing scheme.
It is common that users in a local area share use of a network to access a remote computer or server to realize different services, e.g., data transfer. Because the capacity or bandwidth of the network is usually limited, network management is oftentimes employed to allocate the limited network capacity among the users to realize such services. In obtaining the services, the users typically use their terminals which are connected to a source node of the network to access the server, which is connected to a destination node thereof. In a typical network, the source and destination nodes are connected by multiple links, which aggregately contribute the capacity which is required by the services.
In a prior art network management approach, the shares of the required capacity by the multiple links are determined based on a cost function. For example, the latter is a function of selected network performance objectives and quality of service (QOS) affordable to the users. The actual shares of the required capacity are determined by minimizing such a cost function.
We have recognized that the above-described prior art network management approach in which the cost function minimization dictates the actual shares of the required capacity by the links may not always be desirable, especially when a network provider wants to control such shares based on other criteria concerning, for example, maintaining a certain balance of communication load among the links, minimizing the overall congestion in the network, or reserving capacity of certain links.
The present invention is based on a notion that each service utilizing the links incurs an economic cost to the user requesting the service. That is, the user needs to pay for use of the links for the requested service. In accordance with the invention, the network provider is able to attain a pre-selected target operating point by assessing an appropriate cost to each user for his/her service. The target operating point defines the desired amount of capacity of each link for use. The cost is assessed based on a price which is a function of such a desired capacity amount of each link for use, and the total capacity amount of each link being used. To minimize the cost, each user most likely elects use of the capacity of those relatively inexpensive links to realize a service. Thus, in accordance with the invention, the aforementioned price is dispensed in such a way that it provides users with an incentive to utilize the links in a certain manner, thereby inducing an equilibrium of shares of the required capacity among the links, which coincides with the target operating point.
In the drawing,
FIG. 1 illustrates a communications arrangement in accordance with the invention;
FIG. 2 is a flow chart depicting a process for determining certain factors relating to the cost for using a network in the arrangement of FIG. 1; and
FIG. 3 illustrates a general network to which the invention is also applicable.
Throughout this disclosure, unless stated otherwise, like elements, components or sections in the figures are denoted by the same numeral.
FIG. 1 illustrates communications arrangement 100 embodying the principles of the invention. In arrangement 100, users i, 1≦i ≦N, may respectively utilize terminals 110-1 through 110-N to communicate with server 150, e.g., a host computer, through network 120 to realize different services, e.g., data transfer, where N is an integer greater than or equal to one. Without loss of generality, network 120 includes node 123 to which terminals 110-1 through 110-N are connected, and node 125 to which server 150 is connected. In addition, links 1 through L connect node 123 to node 125, where L is greater than or equal to one. For example, nodes 123 and 125 each may include a conventional router for directing communication flows between server 150 and terminals 110-1 through 110-N through links 1 through L, under the control of network manager 105 described below.
Typically, in realizing services requested by the users, links 1 through L aggregately contribute the bandwidth which is required by the services. In a prior art network management approach, the shares of the bandwidth by links 1 through L are determined by minimizing a cost function which is a function of selected network performance objectives and quality of service (QOS) affordable to the users. However, we have recognized that such a prior art approach in which the cost function minimization dictates the actual shares of the bandwidth by the links may not always be desirable, especially when a network provider wants to control such shares based on other criteria concerning, for example, maintaining a certain balance of communication flows among the links, minimizing the overall congestion in the network, or reserving capacity of certain links.
The present invention is based on a notion that each service utilizing network 120 incurs an economic cost to the user requesting the service. That is, the user needs to pay for use of network 120 for the requested service. For example, such a notion is disclosed in commonly assigned, copending U.S. patent application Ser. No. 09/012,866 filed on Jan. 23, 1998. In accordance with the invention, the network provider is able to attain a pre-selected target operating point F*, which is considered to be desirable, by assessing an appropriate cost to the user for each requested service, depending on, e.g., the availability of each link in the network. This target operating point defines the desired amount of communication flow traversing each link to realize all the services utilizing the network. Thus, in this instance F*=(f1* f2* . . . fL*), where f1* represents the desired amount of communication flow on link 1; f2* represents the desired amount of communication flow on link 2; . . . ; and fL* represents the desired amount of the communication flow on link L.
Specifically, in accordance with a pricing scheme described below, a price is set on a per unit of flow basis, which varies from link to link depending on, among others, the current communication flow thereon. As fully described below, a user is provided with information concerning the cost based on the price on each link for realizing a new service, which requires a certain volume of communication flow over network 120. To minimize the cost, the user tends to elect use of the capacity of those relatively inexpensive links to fulfil the service to which the service provider desires to steer the additional communication flow. Thus, in accordance with the invention, the aforementioned pricing scheme is used to provide the users with an incentive to utilize the links in a certain manner, thereby inducing an equilibrium of the shares of the aggregate flow among the links, which coincides with the aforementioned target operating point.
In order to fully appreciate the pricing scheme in accordance with the invention, we first consider the cost to a user utilizing network 120 to realize a service, which in this instance includes a performance cost and usage cost. The performance cost is based on the QOS provided to the user by each link k, 1≦k≦L, which depends on the current communication flow fk thereon. In this instance, fk may be expressed as follows:
where fk i represents the flow attributed to the service requested by user i, 1≦i≦N, on link k. In addition, the QOS varies with a function Tk indicative of the congestion level on link k. Specifically, the higher the value of Tk, the lower is the QOS. In general, the function Tk is a positive, convex increasing function of fk, signifying that the congestion level increases with the flow thereon, and the link performance deteriorates increasingly as the flow approaches the capacity of the link. In this particular illustrative embodiment, Tk is expressed as the average time delay on link k as follows:
where ck represents the total capacity of link k. In this instance, the quality of service QOSi provided to user i by links 1 through L aggregately may be expressed as follows:
In this instance, such QOSi represents the performance cost to user i.
The usage cost to user i for using the capacity of network 120 to realize the service is determined based on pk, which represents the price for transporting a unit of flow on link k, 1≦k≦L. The unit price pk is a function of the amount of current communication flow fk on link k, but independent of user i. Thus, the service cost Ji to user i consisting of the performance and usage costs for realizing the service may be expressed as follows:
The pricing scheme in accordance with the invention is designed to discourage the users from utilizing congested links. A simple implementation of such a scheme is to require the unit price pk to be proportional to the congestion level of link k. That is:
where φk represents a proportionality constant. Thus, based on expressions  and , the service cost Ji to user i in expression  can be rewritten as follows:
where wk=1+φk. Since the price pk for transporting a unit of flow over link k is always greater than zero, wk must always be greater than zero.
Network manager 105 in FIG. 1, among other things, determines wk for each link k necessary for computation of the service cost according to expression . FIG. 2 illustrates routine 200, which is stored in memory 107 in manager 105, for determining such wk by processor 109. Specifically, routine 200 incorporates an iterative process, with its nth iteration defined as follows:
where fk(n) represents the observed flow currently measured by processor 109 on link k, which corresponds to wk(n) in the nth iteration, and θk represents a predetermined step size which in a well known manner determines the rate of convergence of the iterative process.
Instructed by routine 200, processor 109 initializes wk(n) in the n=0 iteration, e.g., setting wk(0) to 1, as indicated at step 203. Processor 109 at step 207 computes wk(n+1) according to expression . Processor 109 then compares at step 211 fk(n) with the desired amount of flow on link k, i.e., fk*. If fk(n) is not close to fk* within a predetermined tolerance, denoted ε, which is a small number, e.g., 10−5, processor 109 increases n by one, as indicated at step 213, and routine 200 returns to step 207 from there. Otherwise, if fk(n) is close to fk* within ε, processor 109 sets wk to wk(n+1) as indicated at step 215, thereby determining wk.
It should be noted at this point that with wk determined using routine 200, the cost function of expression  gives rise to a “routing game” leading to a well known Nash equilibrium, which in this instance represents the aforementioned target operating point F*. In this routing game, each user attempts to find a routing strategy which minimizes his/her cost. The optimization attempt of the user is affected by the routing strategies of other users. As a result, each user dynamically devises his/her routing strategy subject to the similar action by other users to minimize his/her cost. Such dynamic user actions can be modeled as a non-cooperative game which leads to the Nash equilibrium, from which no user finds it beneficial to deviate unilaterally. By design, the inventive pricing scheme described above causes the Nash equilibrium to coincide with the target operating point. Thus, using the pricing scheme, the target operating point is attained over time when the communication flows on links 1 through L settle.
It should also be noted that based on expression , the aforementioned wk can be envisioned as a cost discount factor of link k, 1≦k≦L. If the service provider increases the discount factor of a particular link, and thus the cost of using that link, the users would reactively move communication flow from such a link to the other links to minimize the service cost. Moreover, it can be shown that when each of the discount factors w1, w2. . . , and wL is multiplied by the same positive number, the Nash equilibrium of the aforementioned routing game remains the same. That is, what determines the characteristics of the routing game is the relative values of wk's. In other words, the relative values of the discount factors are what control the users' behavior in formulating their routing strategies. As a result, the network provider can scale up or down the discount factors simultaneously to adjust the service cost, without affecting the realization of the target operating point.
More importantly, in general, it can be shown that as long as the users are concerned about the relative discount factor values in formulating their routing strategies to minimize the service cost, which is reasonable, the users' actions can always be modeled as a routing game leading to a Nash equilibrium which coincides with a target operating point. The set of discount factors corresponding to the target operating point may be obtained using routine 200 above, regardless of what the actual cost functions of the users are.
Referring back to FIG. 1, when user i, 1≦i≦N, at terminal 110-i requests a new service from server 150 which requires certain capacity, terminal 110-i transmits a request for such a service to network manager 105 through network 120. Upon receiving the request through interface 113, processor 109 determines, among other things, the available capacity, and the aforementioned congestion level Tk and discount factor wk of each link k, 1≦k≦L, and provides information concerning same to terminal 110-i. Based on the provided information, user i or terminal 110-i can determine the allocation of the required capacity among the links, i.e., fk i in the cost function of expression , depending on the user's budget and service requirements such as delay tolerance. Alternatively, processor 109 may determine such an allocation if the service request also includes the user's budget and service requirements. In any event, once the allocation is determined, processor 109 causes nodes 123 and 125 to route the communication flow between terminal 110-1 and server 150 over links 1 through L accordingly to realize the service.
The foregoing merely illustrates the principles of the invention. It will thus be appreciated that a person skilled in the art will be able to devise numerous arrangements which, although not explicitly shown or described herein, embody the principles of the invention and are thus within its spirit and scope.
For example, in the illustrative embodiment, terminals 110-1 through 110-N are connected to server 150 via source node 123 and destination node 125. It will be appreciated that the connections therebetween may be via additional intermediate nodes as illustrated in FIG. 3. For example, such intermediate nodes are denoted 301-1 through 301-M in FIG. 3, where M is an integer greater than or equal to one. Illustratively, the aforementioned set of L links connect node 123 to node 301-1. Similarly, a set of L1 links connect node 301-1 to node 301-2; a set of L2 links connect node 301-2 to node 301-3; . . . ; and a set of LM links connect node 301-M to node 125, where L, L1, L2, . . . and LM may or may not have the same value. In that case, to realize a service from server 150, capacity from one or more links in each set is required. The inventive process of pricing and allocating the required capacity among links 1 through L described before is repeated here for each set, with the service cost associated with each set being additive.
Finally, network manager 105 is disclosed herein in a form in which various system functions are performed by discrete functional blocks. However, any one or more of these functions could equally well be embodied in an arrangement in which the functions of any one or more of those blocks or indeed, all of the functions thereof, are realized, for example, by one or more appropriately programmed processors.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5359593 *||Aug 26, 1993||Oct 25, 1994||International Business Machines Corporation||Dynamic bandwidth estimation and adaptation for packet communications networks|
|US5533108 *||Mar 18, 1994||Jul 2, 1996||At&T Corp.||Method and system for routing phone calls based on voice and data transport capability|
|US6055571 *||Nov 20, 1997||Apr 25, 2000||Nec Usa, Inc.||Computer network with microeconomic flow control|
|US6101180 *||Nov 12, 1997||Aug 8, 2000||Starguide Digital Networks, Inc.||High bandwidth broadcast system having localized multicast access to broadcast content|
|US6134235 *||Oct 8, 1997||Oct 17, 2000||At&T Corp.||Pots/packet bridge|
|US6141325 *||Dec 18, 1996||Oct 31, 2000||International Business Machines Corporation||Paradigm for enabling interoperability between different subnetworks|
|US6230203 *||Mar 14, 1997||May 8, 2001||Scientific-Atlanta, Inc.||System and method for providing statistics for flexible billing in a cable environment|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US6697334 *||Jan 18, 2000||Feb 24, 2004||At&T Corp.||Method for designing a network|
|US6728266 *||Dec 23, 1999||Apr 27, 2004||Nortel Networks Limited||Pricing mechanism for resource control in a communications network|
|US6934259||Nov 12, 2003||Aug 23, 2005||At&T Corp||Apparatus and method for designing a network|
|US7185104 *||Aug 7, 2000||Feb 27, 2007||At&T Corp.||Methods and systems for optimizing network traffic|
|US7188159 *||Aug 9, 2001||Mar 6, 2007||Infineon Technologies Ag||Efficient software download to configurable communication device|
|US7203575 *||Dec 16, 2003||Apr 10, 2007||Rockwell Automation Technologies, Inc.||Decentralized autonomous control for complex fluid distribution systems|
|US7266123 *||Sep 15, 2000||Sep 4, 2007||Intel Corporation||System and method for prioritizing and communicating subscriber voice and data information|
|US7626945||Jul 20, 2005||Dec 1, 2009||At&T Intellectual Property Ii, L.P.||System and method for designing a network|
|US7675890 *||Sep 10, 2004||Mar 9, 2010||Delta Networks, Inc.||QoS based load-balance policy for WLAN|
|US7870233||Feb 15, 2007||Jan 11, 2011||Infineon Technologies Ag||Efficient software download to configurable communication device|
|US7965654||Nov 30, 2009||Jun 21, 2011||At&T Intellectual Property Ii, L.P.||System and method for designing a network|
|US8040892||Aug 9, 2004||Oct 18, 2011||Intel Corporation||Switching system and method for communicating information at a customer premises|
|US8078703||Oct 20, 2010||Dec 13, 2011||Infineon Technologies Ag||Efficient software download to configurable communication device|
|US8149846 *||Nov 9, 2006||Apr 3, 2012||Hewlett-Packard Development Company, L.P.||Data processing system and method|
|US20020004788 *||May 18, 2001||Jan 10, 2002||Gros Thomas D.||Commodity trading of bandwidth|
|US20020045441 *||Aug 9, 2001||Apr 18, 2002||Ralston John D.||Efficient software download to configurable communication device|
|US20040095887 *||Nov 12, 2003||May 20, 2004||At&T Corp.||Apparatus and method for designing a network|
|US20040204784 *||Dec 16, 2003||Oct 14, 2004||Maturana Francisco Paul||Decentralized autonomous control for complex fluid distribution systems|
|US20050013292 *||Aug 9, 2004||Jan 20, 2005||Johnson Brian W.||Switching system and method for communicating information at a customer premises|
|US20050053046 *||Sep 10, 2004||Mar 10, 2005||Shiwei Wang||QoS based load-balance policy for WLAN|
|US20060031082 *||Aug 6, 2004||Feb 9, 2006||Amaitis Lee M||System and method for trading wireless spectrum rights|
|US20060143111 *||Feb 24, 2006||Jun 29, 2006||Cfph, Llc||System and method for trading spectrum rights|
|US20070104099 *||Nov 9, 2006||May 10, 2007||Hewlett-Packard Development Company, L.P.||Data Processing System and Method|
|US20100074115 *||Nov 30, 2009||Mar 25, 2010||At&T Corp.||System and Method for Designing a Network|
|US20100298046 *||May 20, 2010||Nov 25, 2010||Aristocrat Technologies Australia Pty Limited||Gaming system|
|US20110035475 *||Oct 20, 2010||Feb 10, 2011||Infineon Technologies Ag||Efficient software download to configurable communication device|
|US20120033567 *||Jun 7, 2011||Feb 9, 2012||Fujitsu Limited||Data transmission apparatus, data receiving apparatus, and data transmission method|
|U.S. Classification||370/238, 370/411, 370/395.21|
|International Classification||H04L12/14, H04L12/56|
|Cooperative Classification||H04L45/124, H04L12/1414, H04L45/12, H04L45/24, H04L12/1432, H04L12/14, H04L45/123|
|European Classification||H04L45/24, H04L12/14C1, H04L12/14M, H04L45/12, H04L45/124, H04L45/123, H04L12/14|
|Jul 24, 1998||AS||Assignment|
Owner name: LUCENT TECHNOLOGIES, INC., NEW JERSEY
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:AHUJA, SUDHIR R.;KORILIS, IOANNIS A.;REEL/FRAME:009345/0177;SIGNING DATES FROM 19980716 TO 19980721
|Dec 2, 2005||FPAY||Fee payment|
Year of fee payment: 4
|Dec 21, 2009||FPAY||Fee payment|
Year of fee payment: 8
|Dec 19, 2013||FPAY||Fee payment|
Year of fee payment: 12